Non-uniform fog image defogging algorithm based on transmission attention mechanism

A kind of attention and non-uniform technology, applied in the field of image processing, can solve the problems of haze residue, non-uniform haze image and unsatisfactory defogging effect, etc.

Pending Publication Date: 2021-08-20
HENAN POLYTECHNIC UNIV
View PDF5 Cites 3 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

A large number of scholars have applied deep learning to image defogging algorithms and achieved good results, but there are still some shortcomings. For example, the defogging effect on non-uniform haze images is not ideal, and the problem of haze residue is prone to occur.

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Non-uniform fog image defogging algorithm based on transmission attention mechanism
  • Non-uniform fog image defogging algorithm based on transmission attention mechanism
  • Non-uniform fog image defogging algorithm based on transmission attention mechanism

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0044] combine figure 1, a non-uniform fog image defogging algorithm based on a transfer attention mechanism, characterized in that it includes the following steps:

[0045] S1. Sparse and smooth atrous convolution feature extraction, constructing sparse blocks in the form of interleaved ordinary convolution and smooth atrous convolution, extracting feature information of different levels in the image;

[0046] S2. Non-uniform haze feature processing based on the transfer attention mechanism, integrating the attention mechanism into the dehazing network, learning the channel attention feature map and pixel attention feature map in different levels of feature maps, and giving each channel a different The weight value and make the network pay more attention to the dense fog area and high-frequency area, adaptively learn the feature map of the feature map under different weights, and then combine the channel attention feature map and the pixel attention feature map in the attenti...

Embodiment 2

[0050] The overall structure of the transfer attention dehazing network is as follows: figure 2 As shown, the main part of the network is inside the dotted line box. First, the feature information of the input image is initially extracted through three convolutional layers (Conv1~3). The step size of Conv3 is 2, which can reduce the size of the feature map and reduce the network performance. Computational complexity.

[0051] Then, five cascaded sparse blocks (S-1~S-5) are used to extract the features of different layers in the image, avoiding the use of down-sampling operations and reducing the loss of detail information.

[0052] Next, the gated fusion network is used to directly perform gated fusion on the extracted feature maps of the low, medium, and high levels, effectively aggregate image features at different levels, and obtain feature maps containing rich feature information.

[0053] Finally, the fused feature map is restored to the same size as the input fog map b...

Embodiment 3

[0056] Sparse Smooth Atrous Convolutional Feature Extraction.

[0057] The image classification network based on deep learning often expands the receptive field through multi-layer down-sampling operations in order to extract high-level semantic features in the image and reduce the amount of calculation while improving the overall performance of the network. However, the image dehazing network is a reconstruction process at the pixel level , a large number of downsampling operations are easy to lose image detail information, which brings great challenges to the reconstruction of haze-free images. Therefore, it is particularly important to preserve as much detail information as possible while expanding the image receptive field.

[0058] This application proposes a smooth atrous convolution with a sparse structure to achieve feature extraction, and uses gated fusion to fuse features of different levels.

[0059] The sparse mechanism can be effectively applied in the field of im...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention relates to a non-uniform fog image defogging algorithm based on a transmission attention mechanism. The non-uniform fog image defogging algorithm comprises the following steps: S1, performing sparse smooth dilated convolution feature extraction; S2, performing non-uniform haze feature processing based on a transmission attention mechanism; S3, obtaining a loss function. According to the non-uniform fog image defogging algorithm provided by the invention, an end-to-end mapping relation between a degraded image and a clear image is directly constructed under the guidance of an attention mechanism for a real non-uniform haze image, a good defogging effect is achieved for a non-uniform foggy image and a synthesized foggy image, and an obtained restored image has relatively good detail information; the color is more natural.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to a non-uniform fog image defogging algorithm based on a transfer attention mechanism. Background technique [0002] Under foggy conditions, various particles suspended in the air absorb and scatter light, resulting in reduced contrast, color distortion, and blurred edges in the collected outdoor images. The haze images acquired in such environments are neither It is beneficial to the visual observation of images, but also hinders the computer vision tasks in the field of artificial intelligence that use images as the main processing object. Therefore, it is of great research significance and application prospect to study the degradation principle of haze images and improve their clarity. . [0003] Early image defogging algorithms mainly used image enhancement methods to improve the contrast and clarity of the image, and enhance the visual effect of the image...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G06T5/00G06N3/04G06N3/08
CPCG06T5/003G06T5/002G06N3/08G06N3/045
Inventor 王科平杨艺韦金阳李新伟崔立志李冰锋
Owner HENAN POLYTECHNIC UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap